DecksMD 2.0.0

Decks 2.0 — AI flashcards

This release adds an optional AI layer for creating and improving cards, plus a notes field for header-paragraph cards. AI features stay off until you add your own provider key in Settings → AI, and nothing is ever sent to a provider unless you trigger an action.

AI generate

Describe a topic — optionally attaching notes or images — and watch new flashcards stream into an inbox. Keep the ones you want, then save them to a new file or append to an existing deck, in your choice of format: header + paragraph, table, or canvas. The configured profile's header level is used automatically.

AI refactor & split

Rewrite a single card, or batch-refactor a whole selection from the flashcard manager, to improve clarity, fix grammar, and tidy formatting — without changing the meaning. Or split one dense card into several atomic cards. Every suggestion is shown side-by-side and applied only when you accept it. Refactor now returns Markdown (the same readable format the generator uses) instead of JSON.

Notes in header-paragraph cards

Header-paragraph cards can now carry a notes field, like tables. Add a note with an Obsidian comment (%%a hint or mnemonic%%) anywhere in the body, or place it after a --- divider at the end. Notes show on demand (press N) during review.

Bring your own provider

Use OpenAI, Anthropic (Claude), Google (Gemini), or any OpenAI-compatible endpoint (including local servers like Ollama / LM Studio). Keys are stored locally in ai-keys.json and are never written to data.json, so they never leave your device through sync. The request to the provider contains only a built-in prompt describing how Decks cards work, your instructions, and the content for that action.

Localized

The generator, refactor, and AI settings are translated into all supported languages.

Under the hood

The flashcard engine — parsing, FSRS scheduling, sync log, and AI orchestration — now lives in a single shared @decks/core package, removing duplicated code between layers. The per-profile "AI refactor prompt" setting was removed; the built-in prompt teaches the model how Decks cards work, and you can add any extra guidance directly in the AI prompt box.

Existing decks, profiles, and review history are migrated automatically on first launch.